AI use

The one AI feature

AI prioritization takes findings and observations that NEMUQU's own detectors have already produced and asks the model which an analyst should look at first. Parsing, detection, findings, evidence citations, timelines, hunts and reports are produced without AI and work the same with AI turned off.

When it runs

• Only when a member asks for it on a case. Nothing in upload, parsing or detection calls a model.
• Before sending, you can open a preview of the exact payload that will leave NEMUQU.
• Only when the deployment has a provider configured and your plan includes it. Unpaid and retained plans have no AI allowance.

What is sent, and to whom

The request goes to Anthropic, PBC, listed on Subprocessors. It contains the case reference and title, and for each finding its reference, title, narrative, severity, confidence, detector category, ATT&CK technique mapping, benign explanations and unknowns; observation references, titles and details; and the list of evidence reference codes.

Those titles and narratives describe your evidence, so they routinely contain account names, host names, IP addresses and process names. Review the preview with that in mind.

Not sent: your uploaded files, the original log records, evidence values, the entity list and compiled hunt queries.

What comes back, and what it can change

• The model returns references, an order and codes from a fixed list. It has no field in which to write a sentence about your case.
• Every sentence you read next to the ordering is rendered by NEMUQU from its own records, not by the model.
• References that do not exist in the case are rejected and recorded as rejected.
• The model cannot create, edit or delete a finding, observation, hypothesis or action. Its output is an ordering suggestion you are free to ignore.

What is stored

NEMUQU keeps a record of each run: the provider, model, status, token counts, the validated references and codes, any rejected references and a redacted error. The prompt and the raw response are not stored. Run records are part of the case and are deleted with it.

Training and provider retention

NEMUQU has no pipeline that trains models on customer data and does not use customer data to train AI models. What Anthropic retains, for how long, and its own training terms are set by the commercial API agreement and account settings. Those have not been independently verified for this page; ask support@nemuqu.com if your review needs them.

Human oversight

AI prioritization does not decide anything. It does not mark a system compromised or clean, close a case, contact anyone or take a response action. An analyst reads the findings and their evidence and decides what to do. Do not use its ordering as the basis for decisions about individual people, such as employment decisions.

Turning it off

Do not request prioritization, and nothing is sent. There is no background use to opt out of.

Last updated 2026-10-15.